Quasinormal modes and echoes of a double braneworld

Abstract

In this work, we study the gravitational quasinormal modes and the gravitational echoes of a double braneworld. The double braneworld is a kind of split thick brane, which is crucial for addressing the hierarchy problem in the thick brane scenarios. Using the Bernstein spectral method, direct integration method, and asymptotic iteration method, we calculate the quasinormal mode frequencies of the double brane. We find that the quasinormal spectrum is very different from that of the single brane model, especially the high overtone mode. We also perform numerical evolution to study the time-domain properties of the characteristic modes of the double brane. The results show that when the degree of brane splitting is large, gravitational echoes of oscillation attenuation between sub-branes will appear in the thick brane. Furthermore, different long-lived Kaluza-Klein modes interfere with each other, resulting in a beating effect. Compared to a single brane model, the phenomenon of the split double brane is richer, and the lifetime of the massive Kaluza-Klein graviton of the double brane is longer. These phenomena may have potential phenomenological interest. We hope to detect these extra-dimensional signals in future gravitational wave detectors or accelerators.

I Introduction

Braneworld models have been a topic of extensive research for many years, offering a new perspective on the nature of spacetime and providing a novel approach to address the hierarchy problem between the Planck and Electroweak scales ArkaniHamed:1998rs ; Antoniadis:1998ig ; Randall:1999ee . The Randall-Sundrum (RS) models, which involve branes embedded in a five-dimensional anti-de Sitter spacetime, have been particularly influential in this regard. The RS-I model Randall:1999ee proposes a solution to the hierarchy problem by introducing two branes, while the RS-II model Randall:1999vf extends this concept by pushing one brane to infinity. A remarkable feature of the RS-II model is the recovery of the four-dimensional Newtonian potential, even with an infinite extra dimension. These brane models have found wide-ranging applications in various areas of physics, including black hole physics, particle physics, and cosmology Shiromizu:1999wj ; Tanaka:2002rb ; Gregory:2008rf ; Jaman:2018ucm ; Adhikari:2020xcg ; Geng:2020fxl ; Geng:2021iyq ; Geng:2022dua ; Bhattacharya:2021jrn .

While the RS-II thin brane model neglects the thicknesses and internal structure of the brane, treating its energy density as a delta function along the extra dimension, investigating the brane’s inner structure requires considering thick brane models. These models, proposed by DeWolfe et al.DeWolfe:1999cp ; Gremm:1999pj ; Csaki:2000fc , combine the domain wall model without gravityAkama:1982jy ; Rubakov:1983bb with the RS-II model to describe thick branes generated by one or more matter fields. Extensive research has been conducted on thick brane solutions in various gravity theories, as well as the localization of the gravitational zero mode and various matter fields on the brane Afonso:2007gc ; Dzhunushaliev:2010fqo ; Dzhunushaliev:2011mm ; Geng:2015kvs ; Melfo2006 ; Almeida2009 ; Zhao2010 ; Chumbes2011 ; Liu2011 ; Bazeia:2013uva ; Xie2017 ; Gu2017 ; ZhongYuan2017 ; ZhongYuan2017b ; Zhou2018 ; Hendi:2020qkk ; Xie:2021ayr ; Moreira:2021uod ; Xu:2022ori ; Silva:2022pfd ; Xu:2022gth . For more comprehensive information on the braneworld see these reviews Dzhunushaliev:2009va ; Maartens:2010ar ; Liu:2017gcn ; Ahluwalia:2022ttu . In addition to the zero mode, the existence of massive Kaluza-Klein (KK) particles on the brane, which are beyond the standard model. These massive KK modes may not stay on the brane forever, but escape into extra dimensions. Therefore, for these modes, the brane is dissipative. In dissipative systems, these characteristic modes is known as quasinormal modes (QNMs). It has garnered significant attention in various fields, especially in black hole physics Berti:2009kk ; Kokkotas:1999bd ; Nollert:1999ji ; Konoplya:2011qq ; Cardoso:2016rao ; Jusufi:2020odz ; Cheung:2021bol . Recent studies have shown that there may exist a set of discrete modes, known as QNMs of a brane, in both thin and thick braneworld scenarios Seahra:2005wk ; Seahra:2005iq ; Tan:2022vfe ; Tan:2023cra ; Tan:2024url ; Jia:2024pdk ; Tan:2024aym . Investigating these QNMs can provide a deeper understanding of the properties and dynamics of branes. Detecting these modes would open up new ways to understand the nature of spacetime.

One particularly interesting aspect of thick brane models is the phenomenon of brane splitting, where a single thick brane splits into two or more sub-branes Melfo:2002wd ; Castillo-Felisola:2004omi ; deBrito:2014pqa ; deSouzaDutra:2014ddw ; Farokhtabar:2016fhm ; Xie:2019jkq . One fascinating property of the split brane is that it may be possible to solve hierarchical problem simultaneously in the framework of an RS-II model Guerrero:2006gj ; Ahmed:2012nh ; deSouzaDutra:2013rwa . This is the lack of the original RS-II model and the single thick brane models. This scenario offers a rich platform for exploring the properties of thick branes and their associated QNMs. Moreover, the presence of multiple sub-branes in a brane-splitting scenario may give rise to fascinating phenomena, such as gravitational echo. Gravitational echo, which are delayed gravitational wave signals, have been proposed as a potential signature of exotic compact objects and modified gravity theories Cardoso:2017cqb ; Jaramillo:2020tuu ; Witek:2012tr ; Cardoso:2019rvt ; Mark:2017dnq ; Conklin:2017lwb ; Barcelo:2017lnx ; Qian:2024zvq ; Lin:2023qgd . In the context of thick branes, gravitational echoes may emerge due to the interaction of gravitational perturbations with the multiple sub-branes in a brane-splitting scenario Zhu:2024gvl .

In this paper, we aim to investigate the QNMs and potential gravitational echo in a thick brane model with internal structure, focusing specifically on brane-splitting scenarios. By studying the properties of the QNMs, we seek to gain insights into the structure and stability of the brane. Furthermore, we will explore the possibility of gravitational echoes in the brane-splitting scenario. Our findings may offer new perspectives on detecting extra dimensions and probing the nature of gravity in braneworld models.

The rest of this paper is organized as follows. In Sect. II, we introduce the thick brane model with internal structure and brane-splitting, and derive the equations governing the gravitational perturbations. In Sect. III, we use semi-analytic method to solve the quasinormal frequency of the split thick brane in the frequency domain, and further study the QNMs of the thick brane and the properties of the gravitational echo in the time domain by numerical evolution. Finally, the conclusions and discussions are given in Sect. IV.

II Braneworld model in general relativity

In this section, we review the thick brane solution in five-dimensional general relativity briefly and derive the perturbed equation for the transverse-traceless gravitational perturbation. A thick brane can be generated by various matter fields like scalar fields and vector fields. We choose that the thick brane is generated by a canonical scalar field. Thus, the action of the thick brane is

$$S=\int d^{5}x\sqrt{-g}\left(\frac{1}{2\kappa^{2}_{5}}R-\frac{1}{2}g^{MN}\partial_{M}\varphi\partial_{N}\varphi-V(\varphi)\right),$$ (1)

where $\kappa_{5}^{2}=8\pi G_{5}$ is the five-dimensional gravitational coupling constant, which is set to $\kappa_{5}=1$ in this paper. The Einstein equation and the equation of motion for the scalar field are

$$R_{MN}-\frac{1}{2}Rg_{MN} = -\frac{1}{2}g_{MN}\left(\partial^{A}\varphi\partial_{A}\varphi-V(\varphi)\right)$$ (2)

$$+\partial_{M}\varphi\partial_{N}\varphi,$$

$$g^{MN}\nabla_{M}\nabla_{N}\varphi = \frac{\partial V(\varphi)}{\partial\varphi}.$$ (3)

Throughout this paper, indices $M,N,\dots=0,1,2,3,5$ and $\mu,\nu\dots=0,1,2,3$ denote the bulk and the brane coordinates, whereas indices $i,j\dots=1,2,3$ label the three-dimensional space ones on the brane.

By transform to the so-called “gauge” coordinates, a static flat brane metric can be written as Melfo:2002wd

$$ds^{2}=e^{2A(y)}\eta_{\mu\nu}dx^{\mu}dx^{\nu}+e^{2H(y)}dy^{2},$$ (4)

where the four-dimensional Minkowski metric $\eta_{\mu\nu}$ is $\eta_{\mu\nu}=\text{diag}(-1,1,1,1)$. Combining the metric (4), Einstein equation (2), and the equation of motion for the scalar field (3), the specific dynamical equations are

$$2A^{\prime 2}-A^{\prime}H^{\prime}+A^{\prime\prime} = -\frac{1}{2}\varphi^{\prime 2}-V,$$ (5)

$$6A^{\prime 2} = \frac{1}{2}\varphi^{\prime 2}-V,$$ (6)

$$\varphi{{}^{\prime\prime}}+4A^{\prime}\varphi^{\prime} = \frac{\partial V}{\partial\varphi},$$ (7)

where prime is the derivative with respect to $y$. The thick brane solution was investigated in Ref. Melfo:2002wd :

$$A(y) = \delta H(y)=\frac{-\delta}{2}\ln\left(1+\left(\frac{ky}{\delta}\right)^{2s}\right),$$ (8)

$$\varphi(y) = \frac{\sqrt{3\delta(2s-1)}}{s}\arctan\left(\frac{ky}{\delta}\right),$$ (9)

$$V(\varphi) = 3k^{2}\left(\frac{2s+4\delta-1}{2\delta}\cos\left(\frac{s\varphi}{\sqrt{3\delta(2s-1)}}\right)^{2}-2\right)$$ (10)

$$\times\left(\sin\left(\frac{s\varphi}{\sqrt{3\delta(2s-1)}}\right)\right).$$

Here, the dimensionless parameters $\delta$ and $s$ related to the thickness of the brane and the parameter $k$ has mass dimension one. Note that $s$ is a positive odd number. The above solution is generalizations of the thick brane discovered by Gremm et al. Gremm:1999pj . For $s=\delta=1$, the thick brane is returned to the solution in Ref. Gremm:1999pj . In addition, the solution of thick brane will change to RS-II thin brane in $\delta\rightarrow 0$ limit. In this article, we focus on the case of $s>1$ and $\delta\geq 1$. Plots of the warp factor (8), scalar field (9), and scalar potential (10) with different parameters are shown in Figs. 1, 2, and 3. It can be seen that the width of the warp factor increases with the parameters $s$ and $\delta$. From Fig. 2, we can see that the scalar field is a single kink for $s=1$. When $s>1$, the configuration of the scalar field becomes a double-kink, and the scalar field becomes flatter around $y=0$ as $s$ and $\delta$ increase. In addition, for the case of $s>1$, the scalar potential splits at $\phi=0$, which can be seen from Fig. 3. These phenomena indicate that the thick brane splits into two sub-branes when $s>1$, and the splitting degree deepens with the increase of $s$ and $\delta$. This can be clearly seen from the energy density distribution of the thick brane. The energy density of the above thick brane is

$$\rho(y) = -3e^{-2H(y)}\left(A^{\prime\prime}(y)-A^{\prime}(y)H^{\prime}(y)+2A^{\prime}(y)^{2}\right)$$

$$= -\frac{3\delta\left(\left(\frac{ky}{\delta}\right)^{2s}+1\right)^{\frac{1}{s}-2}\left(\frac{ky}{\delta}\right)^{2s}\left(2\delta\left(\frac{ky}{\delta}\right)^{2s}-2s+1\right)}{y^{2}}.$$

We plot the energy density in Fig. 4. It can be seen that, the thick brane splits into two sub-branes for $s>1$, and the distance between the sub-branes increases with the $s$ and $\delta$.

(a)  $\delta=1$

(b)  $s=1$

Figure 1: The shapes of the warp factor (8).

(a)  $\delta=1$

(b)  $s=1$

Figure 2: Plots of the scalar field (9).

(a)  $\delta=1$

(b)  $s=1$

Figure 3: Plots of the scalar potential (10).

(a)  $\delta=1$

(b)  $s=1$

Figure 4: Plots of the energy density (LABEL:energydenisty).

In order to study the characteristic modes of the thick brane, we consider the transverse-traceless tensor perturbation of the thick brane. It is convenient to consider the fluctuation of thick brane in conformal flat coordinates $z$. Introducing the following coordinate transformation $dz=e^{H-A}dy$, the metric (4) can be written as

$$ds^{2}=e^{2A(z)}(\eta_{\mu\nu}dx^{\mu}dx^{\nu}+dz^{2}).$$ (12)

Notice that $dy=dz$ when $\delta=1$. Considering the tensor perturbation of the metric of the thick brane, the perturbed metric can be written as

$$g_{MN}=\left(\begin{array}[]{cc}e^{2A(z)}(\eta_{\mu\nu}+h_{\mu\nu})&0\\ 0&e^{2A(z)}\\ \end{array}\right),$$ (15)

where $h_{\mu\nu}$ is the transverse-traceless perturbation, which satisfies $\partial_{\mu}h^{\mu\nu}=0=\eta^{\mu\nu}h_{\mu\nu}$. Substituting Eq. (15) into the field equation (2), the linear equation for the tensor perturbation is

$$-\frac{1}{2}\Box^{(4)}h_{\mu\nu}-\frac{1}{2}h_{\mu\nu}^{\prime\prime}-\frac{3}{2}A^{\prime}h_{\mu\nu}^{\prime}=0,$$ (16)

where $\Box^{(4)}=\eta^{\alpha\beta}\partial_{\alpha}\partial_{\beta}$. Introducing the following decomposition Seahra:2005iq

$$h_{\mu\nu}=e^{-\frac{3}{2}A(z)}\Phi(t,z)e^{-ia_{j}x^{j}}\epsilon_{\mu\nu},~{}~{}~{}~{}\epsilon_{\mu\nu}=\text{constant},$$ (17)

we can obtain a one-dimensional wave equation of $\Phi(t,z)$ as

$$-\partial_{t}^{2}\Phi+\partial_{z}^{2}\Phi-U(z)\Phi-a^{2}\Phi=0,$$ (18)

where

$$U(z)=\frac{3}{2}\partial_{z}^{2}A+\frac{9}{4}(\partial_{z}A)^{2}$$ (19)

is the effective potential and $a$ is a constant obtained by the separation of variables. Decompose the function $\Phi(t,z)$ further into

$$\Phi(t,z)=e^{-i\omega t}\phi(z).$$ (20)

This yields a Schrödinger-like equation

$$-\partial_{z}^{2}\phi(z)+U(z)\phi(z)=m^{2}\phi(z),$$ (21)

where $m^{2}=\omega^{2}-a^{2}$ denotes the mass of the KK modes. The zero mode of the above equation is

$$\phi_{0}=e^{\frac{3}{2}A(z)}.$$ (22)

In addition, the Schrödinger-like equation (21) can be rewritten as a super-symmetric form

$$QQ^{\dagger}\phi(z)=m^{2}\phi(z),$$ (23)

where $Q$ and $Q^{\dagger}$ are

$$Q=\partial_{z}+\frac{3}{2}\partial_{z}A,~{}~{}~{}~{}~{}~{}~{}~{}Q^{\dagger}=-\partial_{z}+\frac{3}{2}\partial_{z}A.$$ (24)

We can obtain another Schrödinger-like equation with the dual potential as

$$Q^{\dagger}Q\tilde{\phi}(z)=\left(-\partial_{z}^{2}+U^{\text{dual}}(z)\right)\tilde{\phi}(z)=m^{2}\tilde{\phi}(z),$$ (25)

where

$$U^{\text{dual}}(z)=-\frac{3}{2}\partial_{z}^{2}A+\frac{9}{4}(\partial_{z}A)^{2}.$$ (26)

Interestingly, the super-symmetric quantum mechanics guarantees that the Schrödinger-like equations (21) and (25) will share the same massive KK spectrum and QNMs spectrum Cooper:1994eh ; Ge:2018vjq , and that there are no unstable tachyon mode.

III quasinormal modes and echo of the branes

In this section, we investigate the QNMs and the echo of the thick brane. We solve the QNMs of thick brane by the Bernstein spectral method Fortuna:2020obg , the direct integration method Pani:2013pma , and the asymptotic iteration method Cho:2011sf , and study the gravitational echo of thick brane by numerical evolution.

III.1 QNMs of thick brane

First, we use the Bernstein spectral method Fortuna:2020obg to solve the QNMs of the thick brane and compare the results with those obtained by the direct integration method and the asymptotic iteration method. Note that we only consider $\delta=1$, because the coordinate transformation relation between $y$ and $z$ cannot be given analytically when $\delta\neq 1$. Thus, the warp factor $A(z)$, the effective potential $U(z)$, and the dual potential are given by

$$A(z) = \frac{-1}{2}\ln\left(1+\left(\frac{kz}{\delta}\right)^{2s}\right),$$ (27)

$$U(z) = \frac{3k^{2s}z^{2s-2}\left(2-4s+5k^{2s}z^{2s}\right)}{4\left(k^{2s}z^{2s}+1\right)^{2}},$$ (28)

$$U^{\text{dual}}(z) = \frac{3k^{2s}z^{2s-2}\left(k^{2s}z^{2s}+4s-2\right)}{4\left(k^{2s}z^{2s}+1\right)^{2}}.$$ (29)

We plot the above effective potential and the dual ones in Fig. 5. We can see that, the two potentials are volcano-like and the barrier height of both potentials increases with the parameter $s$. When $s=1$, the dual potential is a pure barrier, but when $s>1$, the dual potential splits and a quasi-well appears, which can be seen from Fig. 5(b). The appearance of quasi-well will affect the spectrum of QNMs, and there may be a long-lived resonance mode.

(a)  The effective potential (28)

(b)  The dual potential (29)

Figure 5: The effective potential (28) and the dual potential (29).

We now use the Bernstein spectral method to solve the QNMs of thick brane. First, we briefly review the basic ideas of the Bernstein spectral methods. The Bernstein spectral method is a completely numerical and efficient method for solving the eigenvalues of ordinary differential equations. It has been widely applied to solve the QNMs of black holes. Consider a linear differential equation with a linear differential operator $\hat{L}(u,\omega)$

$$\hat{L}(u,\omega)\Phi(u)=0,$$ (30)

where $\omega$ is the eigenvalue and $u\in[a,b]$ is a compact coordinate. Using the Bernstein basis, the function $\Phi(u)$ is expanded to

$$\Phi(u)=\sum_{k=0}^{N}C_{k}B_{k}^{N}(u),$$ (31)

where $C_{k}$ is expansion coefficient and

$$B_{k}^{N}=\frac{N!}{k!(N-k)!}\frac{(u-a)^{k}(b-u)^{N-k}}{(b-a)^{N}}$$ (32)

is the Bernstein polynomial basis. Substituting Eq. (31) into Eq. (30) with appropriate boundary conditions, the following generalized eigenvalue equation for the set of coefficients C can be obtained

$$\textbf{M}(\omega)\textbf{C}=0.$$ (33)

By solving the above equation (33), we can get the eigenvalue of the differential equation (31). More details on the Bernstein spectral methods can be found in Ref. Fortuna:2020obg . Next, we use the Bernstein spectra method to solve the QNMs of the thick brane.

The specify form of the Schrödinger-like equation (21) for the effective potential (28) is

$$-\partial_{z}^{2}\phi(z)+\left(\frac{3k^{2s}z^{2s-2}\left(2-4s+5k^{2s}z^{2s}\right)}{4\left(k^{2s}z^{2s}+1\right)^{2}}-m^{2}\right)\phi(z)=0.$$

Since the massive KK modes escapes to infinity of extra dimension, the boundary conditions should be

$$\tilde{\phi}(z)\propto\left\{\begin{aligned} e^{imz},&~{}~{}~{}~{}~{}z\to\infty.&\\ e^{-imz},&~{}~{}~{}~{}~{}z\to-\infty.&\end{aligned}\right.$$ (35)

Because the Bernstein spectral method is effective in the compact coordinate domain, but here the extra dimensions are infinite. The Bernstein spectrum method cannot be used directly. We need to transform our coordinates to be finite. So we introduce the transformation $u=\frac{\sqrt{4k^{2}z^{2}+1}-1}{2kz}$. Now, Eq. (III.1) can be rewritten as

$$\left(\frac{m^{2}}{k^{2}}-\frac{3\left(\frac{u}{1-u^{2}}\right)^{2(s-1)}\left(\left(\frac{u}{1-u^{2}}\right)^{2s}+4s-2\right)}{4\left(\left(\frac{u}{1-u^{2}}\right)^{2s}+1\right)^{2}}\right)\phi(u)$$

$$+\frac{\left(u^{2}-1\right)^{3}\left(\left(u^{4}-1\right)\phi^{\prime\prime}(u)+2u\left(u^{2}+3\right)\phi^{\prime}(u)\right)}{\left(u^{2}+1\right)^{3}}=0,$$

where $u\in[-1,1]$. The boundary conditions (35) become

$$\phi(u)\propto\left\{\begin{aligned} e^{-\frac{im/k}{2u-2}},&~{}~{}~{}u\to 1.&\\ e^{\frac{im/k}{2u+2}},&~{}~{}~{}u\to-1.&\end{aligned}\right.$$ (37)

Further rewrite the function $\phi(u)$ as

$$\phi(u)=\psi(u)e^{-\frac{im/k}{2u-2}}e^{\frac{im/k}{2u+2}}.$$ (38)

Substituting the above expression (38) into Eq. (III.1), we have a final equation that is suitable for the Bernstein spectral method

$$s_{1}(u)\psi(u)+s_{2}(u)\psi^{\prime}(u)+s_{3}(u)\psi^{\prime\prime}(u)=0,$$

where

$$s_{1} = 4\left(u^{2}-1\right)^{2}\left(u^{2}+1\right),$$ (39)

$$s_{2} = \frac{8u\left(k\left(u^{4}+2u^{2}-3\right)+2im\left(u^{2}+1\right)\right)}{k},$$ (40)

$$s_{3} = \frac{4m^{2}}{k^{2}}\left(u^{2}+1\right)-\frac{8im}{k}\left(u^{2}-1\right)$$

$$-\frac{3\left(u^{2}+1\right)^{3}\left(\left(\frac{u}{1-u^{2}}\right)^{2s}+4s-2\right)\left(\frac{u}{1-u^{2}}\right)^{2s}}{u^{2}\left(\left(\frac{u}{1-u^{2}}\right)^{2s}+1\right)^{2}}.$$

We use the Mathematica package Fortuna:2020obg which is called SpectralB to obtain the QNMs of the thick brane. The results are compared with those obtained by the direct integration method, which are shown in Tab. 1. The effect of the parameter $s$ on the first three quasinormal frequencies (QNFs) is shown in Fig. 6. From Tab. 1 and Fig. 6, we can see that the real parts of the second and third QNFs and the imaginary parts of the first three QNFs increase with the parameter $s$. The real part of the first QNF first increases with $s$ and then decreases. This shows that the lifetime of the first three QNMs on the brane increases with $s$. It can also be seen that when $s$ changes from 1 to 3, the first three QNFs change more dramatically. Then, with the increase of $s$, the first three QNFs change gently. This is because when $s$ from 1 to 3 the thick brane splits into two sub-brane, and the structure of the thick brane completely changes. Then $s$ increase only gradually changes the thickness of each sub-brane and the distance between sub-branes, so the quasinormal spectrum changes more gently.

$\;\;s\;\;$	$\;\;n\;\;$	Bernstein spectral method	Direct integration method	Time evolution
		$\text{Re}(m/k)$    $\text{Im}(m/k)~{}~{}$	$~{}~{}~{}~{}~{}~{}\text{Re}(m/k)$    $\text{Im}(m/k)~{}~{}$	$~{}~{}~{}~{}~{}~{}\text{Re}(m/k)$    $\text{Im}(m/k)~{}~{}$
1	1	0.99702   -0.52636	0.99702    -0.52636	0.99117     -0.52755
3	1	1.04138   -0.19783	1.04138    -0.19783	1.04012     -0.19778
	2	2.15785   -0.88434	2.15785    -0.88434	non        non
5	1	1.03749   -0.17009	1.03749    -0.17009	1.03666     -0.17000
	2	2.27955   -0.67896	2.27955    -0.67896	non        non
7	1	1.03521   -0.16249	1.03521    -0.16249	1.03449     -0.16240
	2	2.31570   -0.61571	2.31570    -0.61571	non        non
9	1	1.03395   -0.15933	1.03395    -0.15933	1.03327     -0.15923
	2	2.33110   -0.58903	2.33110    -0.58901	non        non
11	1	1.03321   -0.15770	1.03321    -0.15770	1.03254     -0.15760
	2	2.33898   -0.57558	2.33896    -0.57539	non        non
13	1	1.03274   0.15676	1.03273    -0.15676	1.03208     -0.15666
	2	2.34363   -0.56831	2.34347    -0.56754	non        non

Table 1: The frequencies of low overtone modes with different $s$ by the Bernstein spectral method, the direct integration method, and the time evolution.

(a)  The first QNF

(b)  The first QNF

(c)  The second QNF

(d)  The second QNF

(e)  The third QNF

(f)  The third QNF

Figure 6: The relation between the real (left panel) and the imaginary (right panel) parts of the first three QNFs and the parameter $s$.

In addition, we also investigated the high overtone modes of the thick brane. The results are shown in Tabs. 2, 3, and Fig. 7. From Fig. 7(a) we can see that the real part of the first ten QNFs decreases with the overtone number $n$. This is very similar to the Schwarzschild black hole situation. Because the dual potential of the thick brane when $s=1$ is similar to the shape of the effective potential of the Schwarzschild black hole. But when the brane splits, the situation is very different. Table 3 and Fig. 7(b) show the first ten QNFs of thick brane with different parameters $s$. From Fig. 7(b) we can clearly see that, in contrast to the case of $s=1$ (the brane does not split), the real part of the first ten QNFs increases with $n$. Therefore, the splitting of thick brane has a great influence on its quasinormal spectrum.

On the other hand, these results also demonstrate the stability of low overtone modes. As can be seen from Fig. 7(b), when the parameter $s$ changes, the modes with small absolute value of imaginary part, especially the modes with the smallest imaginary part (the fundamental mode), changes little. On the contrary, higher overtone modes vary more. This is the same as the result indicated recently by the black hole pseudo-spectrum Jaramillo:2020tuu , that is, the high overtone modes are highly sensitive to changes in the effective potential, while the fundamental mode is stable. In general, the results in the above frequency-domain show that the quasinormal spectrum of the split thick brane is completely different from that of the unsplit thick brane. In order to study the characteristic mode of the split thick brane more completely, we will perform the time-domain analysis in the next section.

$\;\;s\;\;$	$\;\;n\;\;$	Bernstein spectral method	Asymptotic iteration method
		$\text{Re}(m/k)$    $\text{Im}(m/k)~{}~{}$	$~{}~{}~{}~{}~{}~{}\text{Re}(m/k)$    $\text{Im}(m/k)~{}~{}$
	1	0.99702   -0.52636	0.99702    -0.52636
	2	0.58371   -1.85156	0.58371    -1.85156
	3	0.35314   -3.54785	0.35271    -3.54771
1	4	0.28388   -5.20906	0.28290    -5.20962
	5	0.25268   -6.83448	0.24770    -6.83235
	6	0.23631   -8.44131	0.23933    -8.43357
	7	0.22560   -10.04081	0.21571   -10.05330
	8	0.21927   -11.63293	0.21294   -11.64998

Table 2: The first eight QNFs using the Bernstein spectral method and the asymptotic iteration method.

$\;\;s\;\;$	$\;\;n\;\;$	Bernstein spectral method	Direct integration method
		$\text{Re}(m/k)$    $\text{Im}(m/k)~{}~{}$	$~{}~{}~{}~{}~{}~{}\text{Re}(m/k)$    $\text{Im}(m/k)~{}~{}$
	1	1.03203   -0.15550	1.03203    -0.15547
	2	2.35270   -0.56578	2.34944    -0.55697
	3	3.85612   -0.82156	3.87348    -0.84437
	4	5.43447   -1.03258	5.42952    -1.05052
19	5	6.96103   -1.23800	6.99144    -1.22080
	6	8.54970   -1.39227	8.55497    -1.37331
	7	10.09836   -1.53539	10.11913   -1.51628
	8	11.67300   -1.64354	11.68364   -1.65388
	9	13.27121   -1.78436	13.24841   -1.78836
	10	14.79788   -1.92809	14.81335   -1.92101
	1	1.03180   -0.15534	1.03191   -0.15526
	2	2.35691   -0.57003	2.35037   -0.55528
	3	3.85274   -0.80636	3.87600   -0.83896
	4	5.43948   -1.01395	5.43365   -1.03948
21	5	6.95075   -1.22743	6.99710   -1.20254
	6	8.55633   -1.37537	8.56201   -1.34654
	7	10.09980   -1.50853	10.12741   -1.47998
	8	11.67796   -1.59551	11.69306   -1.60725
	9	13.28684   -1.72391	13.25887   -1.73083
	10	14.80613   -1.86478	14.82482   -1.85214

Table 3: The first ten QNFs using the Bernstein spectral method and the direct integration method.

(a)  $s=1$

(b)  $s>1$

Figure 7: The first ten quasinormal frequencies of the brane solved by the Bernstein spectral method with different $s$.

III.2 Time domain: quasinormal and echo waveforms

We have studied the QNMs of the thick brane in the frequency-domain by various methods. These results show how these fluctuations oscillate and decay when the thick brane is perturbed. But this is not complete, only in the frequency-domain research lacks a description of the evolution of these fields waveform, the relative amplitude, etc. Therefore, it is necessary to consider the evolution of the $1+1$ wave equation (18). First, we still consider the case of $\delta=1$, where the effective potential has an analytical expression in $z$ coordinate. We can easily change the coordinates to the light-cone coordinates $u=t+z$ and $v=t-z$ for evolution. Then evolution equation (18) can be written as

$$\left(4\frac{\partial^{2}}{\partial u\partial v}+U+a^{2}\right)\Phi(u,v)=0.$$ (42)

We consider firstly the initial data is a Gauss packet with width $\sigma$ and located at $v_{c}$

$$\Phi(0,v)=e^{\frac{-(v-v_{c})^{2}}{2\sigma^{2}}},~{}~{}~{}\Phi(u,0)=e^{\frac{-v_{c}^{2}}{2\sigma^{2}}},$$ (43)

We chose the Gaussian wave packet with width $k\sigma=1$ and located at $kv_{c}=20$. We set the parameter $a$ to $a/k=1$. The evolution waveform of Gauss pulse is shown in Fig. 8, which show how the initial wave packet evolves over time at different locations in the extra dimension. It can be seen that after the initial oscillation damping, the evolution of the wave packet tends to be a sinusoid. The frequencies of these final sinusoids does not depend on the extraction point $z_{\text{ext}}$, but the amplitude decreases with $z_{\text{ext}}$. This is due to the existence of a bound zero mode of the thick brane, which does not decay. To illustrate this more clearly, we extract the late maximum amplitude of the evolved waveform at different points and compare it with the analytic zero mode (22), the results are shown in Fig. 9. It can be seen that the profile obtained by the numerical evolution is consistent with the shape of the analytic zero mode.

(a)  $kz_{\text{ext}}=0$

(b)  $kz_{\text{ext}}=5$

(c)  $kz_{\text{ext}}=10$

Figure 8: Evolution waveforms of the Gauss packets (43) at different locations. The data are extracted at the points $kz_{\text{ext}}=0,5,10$.

Figure 9: Compare the zero mode (dot) exited by the Gauss pulse with the analytical zero mode (22) (line).

However, because the information of the QNMs is covered by the zero mode, it is difficult to obtain the information of the QNMs of the thick brane from the evolution of the Gaussian wave packet. There are two ways to solve this problem, one is to place the extraction point far away from the zero point of the extra dimension $z$, then the effect of the zero mode becomes very weak. But that requires a lot of computing resources. Another method is given by the symmetry of the brane. Because of the $Z_{2}$ symmetry, the KK mode can only be odd or even. It is clear that the zero mode is an even function, and the first excited state, the first QNM, is odd-parity. Therefore, when we consider that the initial wave packet is static and odd-parity, only the odd-parity QNMs will be excited, so that we can study the evolution waveform of odd-parity QNMs. The form of a static odd wave packet is as follows

$$\Phi(0,v)=\sin\left(\frac{kv}{2}\right)e^{\frac{-k^{2}v^{2}}{4}},$$

$$\Phi(u,0)=\sin\left(\frac{ku}{2}\right)e^{\frac{-k^{2}u^{2}}{4}}.$$ (44)

We plot the waveforms for different $s$ in Fig. 10. Note that we set $a=0$ for this case. It can be seen that, after the initial burst phase, the wave packet evolution has the characteristic of a typical QNM: the wave packet first exponentially decay, and then appears a power-law damping tail at a late stage. It can also be seen from Fig. 10 that with the increase of $s$, the damping of the initial wave packet slows down, which is consistent with the results obtained in the frequency domain. In fact, the frequency of the first QNM can be obtained by fitting the evolutionary data. Table 1 also lists the frequency of the first QNM of the thick brane obtained by numerical evolution for different parameters $s$, which is consistent with the results obtained in the frequency domain. These results show that, for thick brane, there are discrete characteristic modes in continuous massive KK modes. The properties of these characteristic modes depend on the configuration of the extra dimensions.

Figure 10: Evolution waveform of the static odd packet (44) at $kz_{\text{ext}}=30$ in logarithmic scale.

Now, let us turn our attention to another interesting phenomenon: the echo. When $s>1$, the thick brane splits, and the dual potential results in a quasi-well. The quasi-well is similar to a resonant cavity; thus, the waves will be quasi-bound in the cavity and slowly leak out, resulting in an echo. In the previous part, we did not observe the echo because the two sub-barriers of the effective potential, or the two sub-branes of the thick brane, are too close to support the presence of the echo signal. But when we adjust $\delta$, things change. We plot the shapes of the effective potential and energy density corresponding to different $\delta$ with $s=35$ in Fig. 11. It can be seen that the distance between the two sub-barriers of the effective potential and the peaks of the energy density increases with $\delta$, indicating that the degree of brane splitting intensifies.

(a)  The effective potential $U(z)$

(b)  the energy density $\rho(z)$

Figure 11: The effective potential and the energy density for different $\delta$ with $s=35$.

We now consider the evolution of Gaussian wave packets on these thick branes. We focus on the Gaussian wave packet incident to the left from the right peak of the energy density to simulate the emission of gravitational waves from one sub-brane to another. Figure 12 shows the evolution waveform of the incident Gaussian wave packet on the brane over time at different $\delta$ observed at the zero point of the extra dimension. We can see that, for the parameter we selected $\delta$, the echo signal appears in the time-domain evolution waveform. When $\delta$ is small, the echo signal is masked by the zero mode after some time, as shown in Fig. 12(a). As $\delta$ increases further, the echo signal becomes apparent, but the overall evolutionary waveform seems to become less regular. This happens in the evolution of a massive field in a black hole background and is known as the beating effect Witek:2012tr . This phenomenon can be explained by the beat modulation between two or more long-lived modes with similar frequencies. This beat modulation depends on the relative intensity of the different overtones, which in turn depends on the extraction point, as in the case of guitar strings, a specific mode cannot be excited at its nodes. These characteristics should be universal as long as there are at least two long-lived modes present.

(a)  $s=35$, $\delta=5,kv_{c}=5$

(b)  $s=35$, $\delta=10,kv_{c}=11$

(c)  $s=35$, $\delta=15,kv_{c}=16$

Figure 12: Evolution waveform of the Gauss packet (43) at $kz_{\text{ext}}=0$ for different $\delta$. The parameter of the Gaussian wave packet is $k\sigma=1$.

To understand the beating effect more directly, it is useful to observe the waveform evolution of different extraction points and the corresponding Fourier spectrum. In Fig. 13, we show the waveforms and corresponding Fourier spectrum of Gaussian wave packets evolving over time at different extraction points when $\delta=10$ and $s=35$. The extraction points are selected as $kz_{\text{ext}}=10$ (the location of the sub-brane where the gravitational wave occurs), $kz_{\text{ext}}=-10$ (the location of another sub-brane) and $kz_{\text{ext}}=50$ (the simulated infinity of extra dimensions). These results clearly show that several peaks correspond to different mode contributions of KK gravitons (zero mode or long-lived quasi-normal modes), and the relative amplitudes of these peaks vary significantly as the extraction point changes. This confirms our previous analysis of the beating effect.

In addition, we also briefly consider the effect of the parameter $a$, the results are shown in Fig. 14. It can be seen that the oscillation frequency of the evolving waveform increases with $a$. This is due to the oscillation frequency $\omega^{2}=a^{2}+m^{2}$. The above numerical evolution results show that when the degree of brane splitting is high, the oscillation of the massive KK mode on the brane will indeed echo. However, the existence of zero mode and beating effect masks a lot of information, and it is difficult to obtain information such as the period of the echo.

(a)  $s=35$, $\delta=10$, $kz_{\text{ext}}=10$

(b)  $s=35$, $\delta=10$,$kz_{\text{ext}}=10$

(c)  $s=35$, $\delta=10$, $kz_{\text{ext}}=-10$

(d)  $s=35$, $\delta=10$, $kz_{\text{ext}}=-10$

(e)  $s=35$, $\delta=10$, $kz_{\text{ext}}=50$

(f)  $s=35$, $\delta=10$, $kz_{\text{ext}}=50$

Figure 13: Left panel: Evolution waveform of the Gauss packet (43) at different locations $kz_{\text{ext}}$. Right panel: Corresponding spectrum obtained by the Fourier transformation. The red dashed vertical lines correspond from left to right to the oscillation frequency of the zero mode, first QNM, second QNM, and so on. The parameters of Gauss packet (43) are $kv_{c}=11$ and $k\sigma=1$.

(a)  $s=35$, $\delta=10$, $a/k=1/2$

(b)  $s=35$, $\delta=10$, $a/k=3/2$

Figure 14: Evolution waveforms of the Gauss packet (43) at $kz_{\text{ext}}=0$ for different $a$. The parameters of the Gauss packet (43) are set to $kv_{c}=11$ and $k\sigma=1$.

Thus, we consider the evolution of the static odd-parity wave packet that excludes the effects of zero mode. The wave packet is also chosen in the form of Eq. (44). Figure 15 shows the evolution of odd-parity wave packets with different $\delta$. We can see that when $\delta$ is large, the wave packet oscillates in the quasi-well to produce an echo, and the amplitude of the echo decreases with time. The period of the echo can be clearly observed in the corresponding semi-log coordinate. It can be seen that when $\delta=15$, the echo period is about 30, and when $\delta=25$, the echo period is about 50, which is exactly the same as the width of the corresponding potential well.

On the other hand, in the case of odd-parity initial data, we do not seem to observe a beating phenomenon. To verify this, we select different extraction points to observe the evolution waveform, and the results are shown in Fig. 16. It can be seen that the shape of waveforms at different extraction points is basically the same, which indicates that there is indeed no beating effect under the odd-parity initial data when $a=0$. This may be because the odd-parity wave packet only excites the odd QNMs, so it does not interfere with the zero mode. Additionally, the frequency gap between the different odd QNMs is large enough that they do not affect each other.

According to the above results, the split thick brane exhibits a richer phenomenology, including changes in the spectral structure of QNMs, the emergence of long-lived QNMs, echoes, and the beating effect. For the case of $\delta=1$, these phenomena should be extremely weak because the lifetime of the QNMs in this case is still extremely short compared to that of the bound zero mode. For example, in the case of $s=21$ and $\delta=1$, the lifetime of the first QNM is about $10^{-12}$ seconds when $k=10^{-3}$eV. However, for the case of $\delta>1$, as the degree of brane splitting increases, the lifetime of the QNMs of the thick brane will also increase, which may have a considerable degree of impact on the brane. We hope to search for these extra-dimensional messengers in colliders or gravitational wave detectors.

(a)  $s=35,\delta=5$

(b)  $s=35,\delta=5$

(c)  $s=35,\delta=15$

(d)  $s=35,\delta=15$

(e)  $s=35,\delta=25$

(f)  $s=35,\delta=25$

Figure 15: Left panel: Evolution waveforms of the static odd-parity wave packet (44) at $kz_{\text{ext}}=5$ for different values of $\delta$. Right panel: Same as the left panel but plotted on a semi-log scale.

(a)  $s=35$, $\delta=15$, $kz_{\text{ext}}=15$

(b)  $s=35$, $\delta=15$, $kz_{\text{ext}}=50$

Figure 16: Evolution waveforms of the static odd packet (44) at different points $kz_{\text{ext}}$.

IV Conclusion and discussion

In this paper, we investigated the gravitational QNMs and echoes of the split thick brane. The splitting of the brane results in a richer phenomenology. The quasinormal spectrum of the split thick brane is very different from that of the unsplit ones, especially the high overtone mode. As the degree of splitting increases, echo signals and a beating effect appear, which are unique features of thick branes with sub-brane structures.

We first review the split thick brane model and the corresponding transverse-traceless tensor fluctuations. We then derive the wave equation and the Schrodinger-like equation along the extra dimension for spin-2 KK gravitons on the brane. Based on the above equations, we study the quasinormal spectrum of the split thick brane in the frequency domain by the Bernstein spectral method, asymptotic iteration method, and direct integration method. The results are shown in Figs. 6, 7, Tabs. 1,  2, and  3. We find that the spectrum for $s>1$ (corresponding to a split brane) is completely different from that for $s=1$ (corresponding to an unsplit brane). However, when $s$ continues to increase, the fundamental mode changes very little, but the high overtone modes change greatly. This indicates that the fundamental mode is relatively stable to the change of potential, while the high overtone modes are extremely sensitive to the change of potential, suggesting that the high overtone modes may be the key to identify the thick brane with little difference in configuration. To complement the results in the frequency domain, we also perform time-domain evolution. The evolution results are shown in Figs. 8 and 10. The frequency of the first QNM obtained by numerical evolution is consistent with the results in the frequency domain. Finally, we study the gravitational echo of the thick brane with $\delta>1$. The results show that echo signals appear for larger values of $\delta$. This is because with the increase of $\delta$, the splitting degree of the thick brane also increases, the effective potential is similar to a resonant cavity, and the gravitational KK mode oscillates and leaks in the quasi-well, forming an echo. In addition, we also found that the different KK modes of the thick brane interfered with each other, resulting in the beating effect. This lays a foundation for further research on the co-evolution of multiple KK modes on the brane. In general, the splitting of the brane enriches the structure of extra dimensions and also leads to more interesting phenomenology. We hope to detect these extra-dimensional signals, especially echo signals, in future gravitational wave detection experiments.

Our work could be enhanced by several ways. For example, considering thick brane models with more sub-branes or brane arrays. The characteristic model of thin brane model with sub-brane structure is also worth studying. Characteristic modes and echoes for other test fields should also be investigated.